Pytorch implements the calculation of flops and params

Concept

• FLOPS： Refers to the number of floating-point operations per second , It means computing speed , Is a measure of hardware performance .

 Calculation formula ：
 For convolution ：(K_h * K_w * C_in * C_out) * (H_out * W_out)
 For the full connectivity layer ：C_in * C_out

• FLOPs： Be careful s A lowercase letter , Refers to a floating-point operand , It means the amount of calculation . It can be used to measure the algorithm / The complexity of the model .
• Params： It refers to the total number of parameters to be trained in model training .

 The calculation formula of model parameters is ：
 For convolution ：（K_h * K_w * C_in）* C_out
 For the full connectivity layer ：C_in * C_out

Code implementation

Method 1

Use thop library

pip install thop

import torch
from torchvision import models
from thop import profile


if __name__ == '__main__':
    net = models.mobilenet_v2()
    inputs = torch.randn(1, 3, 224, 224)
    flops, params = profile(net, inputs=(inputs, ))
    print("FLOPs=", str(flops/1e9) +'{}'.format("G"))
    print("params=", str(params/1e6)+'{}'.format("M"))

Method 2

Use torchstat library

pip install torchstat

import torch
from torchvision import models
from torchstat import stat

if __name__ == '__main__':
    net = models.mobilenet_v2()
    stat(net, (3, 224, 224))